Scoring was not performed

• Migrate from BAML-generated tool definitions to explicit MCP ToolDescription registration. Register each tool with complete JSON Schema, descriptions for all parameters and nested properties, and documented return types.
• Add output schema documentation for each tool. Specify the exact structure of CodeSnippet, ranked candidate objects from MatchCapability, and categorized capability objects from SummarizeCapabilities.
• Enhance parameter descriptions: for 'tool' param in GenerateCodeSnippet, describe what ToolInfo structure must contain. For 'manifest' in MatchCapability, document the expected format of each server/CLI entry and what fields are required vs optional.
• Implement error handling patterns. Wrap tool logic in try-catch blocks that return structured error objects with recovery guidance. E.g., 'No matching capabilities found. Try rephrasing the query or ensure the manifest includes relevant servers.'
• Optimize tool descriptions for LLM selection. Rewrite each description as: WHAT it does, WHEN to use it (vs. similar tools), and WHAT it returns. Example: 'Matches user-requested capabilities against available MCP servers and CLI tools, returning ranked candidates by relevance. Use this to help users discover which tools can satisfy their request.'
• Add input validation and constraints. For 'query' in MatchCapability, specify minimum/maximum length and character restrictions. For 'available_clis' array, document expected format (string identifiers).
• Document dependencies and prerequisites. Explain that GenerateCodeSnippet requires tool_name, description, and args array to produce usable code. Clarify that MatchCapability requires a valid manifest structure to function.